WEBVTT 00:09.320 --> 00:10.040 Hi everyone. 00:10.040 --> 00:11.520 Welcome to Steam Academy. 00:11.560 --> 00:12.440 My name is Avril. 00:12.920 --> 00:17.640 Today we are going to talk about what is crosstalk in a transmission line in terms of aggressor and 00:17.640 --> 00:21.280 victim nets and what are the sources of crosstalk in a design. 00:21.280 --> 00:26.240 Then we will discuss what is near end and far end crosstalk, and how to reduce it in a transmission 00:26.240 --> 00:30.880 line, along with a couple of simulations in security Aurora 17.4 and Topology Explorer. 00:31.040 --> 00:36.200 So let's get started. 00:36.360 --> 00:40.280 We will start with first question what is crosstalk in a transmission line? 00:40.640 --> 00:47.600 As its name suggests, crosstalk is transfer of an unwanted signal from one net to adjacent net, and 00:47.600 --> 00:52.440 it occurs between every pair of net which are at close proximity of each other. 00:53.520 --> 00:57.240 The term net we are using for signals and return path both. 00:57.240 --> 01:05.010 We will understand that in more detail while discussing ground Now here we are using few terms for the 01:05.010 --> 01:05.970 source of noise. 01:05.970 --> 01:08.810 We call the aggressor net or active net. 01:09.250 --> 01:15.050 For the net on which noise generated due to crosstalk, we call them quiet or victim nets. 01:15.290 --> 01:16.930 As you can see on your screen. 01:21.770 --> 01:28.090 Now let's talk about what is the origin of crosstalk, how this coupling occurs between two parallel 01:28.090 --> 01:33.570 running transmission lines, and what is the role of fringe field which cause capacitive and inductive 01:33.570 --> 01:34.050 coupling? 01:35.010 --> 01:36.890 First we will talk about fringe field. 01:36.890 --> 01:39.250 If you can recall from return path video. 01:39.650 --> 01:44.970 There I have explained how electric field and magnetic field line is spread between signal and return 01:44.970 --> 01:45.970 path conductors. 01:46.770 --> 01:50.250 These field lines are not just confined to return path. 01:50.290 --> 01:53.730 This spread of field lines affect the surrounding conductors as well. 01:53.850 --> 01:55.490 We call them fringe fields. 01:56.370 --> 02:02.020 As you can see on the screen, if we put quiet net closer to active net, the fringe feel effect will 02:02.020 --> 02:06.980 be more, but it will increase the pitch between these two parallel running tracks. 02:07.140 --> 02:09.740 The effect of fringe field will reduce. 02:10.660 --> 02:16.580 So we can say that the effect of fringe field is inversely proportional to pitch between parallel running 02:16.620 --> 02:17.100 tracks. 02:17.980 --> 02:24.100 The only way noise will be picked up by victim net or quiet line is when there is change in voltage 02:24.100 --> 02:26.540 or current in aggressor or active net. 02:26.980 --> 02:32.420 And that change in voltage and current cause change in electric and magnetic field, which will induce 02:32.420 --> 02:34.060 current in adjacent conductor. 02:35.380 --> 02:41.860 So as a result these fringe field arise small capacitance and inductance between two parallel running 02:41.900 --> 02:42.380 tracks. 02:42.780 --> 02:45.700 And we call them mutual capacitance and mutual inductance. 02:47.100 --> 02:47.540 Solution. 02:47.540 --> 02:54.100 To avoid fringe field noise or to reduce mutual capacitance and inductance, we have to place nets far 02:54.100 --> 02:54.940 from each other. 03:00.780 --> 03:06.400 In the next step, I'll give you brief about the simulation of crosstalk in Aurora 17.4. 03:06.680 --> 03:10.440 And then we will talk about concept of near and far end crosstalk. 03:12.000 --> 03:14.280 So let's open security Aurora 17.4. 03:14.360 --> 03:16.600 And here I have done the simulation already. 03:17.200 --> 03:23.160 As you can see I have simulated for one byte group from Parallel Data Bus zero to Parallel Data Bus 03:23.160 --> 03:23.640 seven. 03:24.720 --> 03:29.080 And the simulation highlighted all the part which has maximum crosstalk. 03:29.360 --> 03:33.920 And as you can see, parallel data seven and parallel data five are highlighted. 03:34.520 --> 03:36.840 We have simulated for parallel data bus seven. 03:36.840 --> 03:43.480 So that means parallel data seven is the victim net and parallel data five is the aggressor net or active 03:43.480 --> 03:43.840 net. 03:44.760 --> 03:47.880 From this plot also we can verify the parallel data. 03:47.920 --> 03:53.160 Bus seven has maximum crosstalk of 110.3 millivolt. 03:54.760 --> 04:00.960 And if I plot for parallel data five as well to see why there is a crosstalk, you can see there is 04:00.960 --> 04:08.850 a clear transition of 1 to 0, and due to that there is a current induced on parallel running tracks. 04:09.010 --> 04:09.850 Data seven. 04:10.130 --> 04:12.570 Similarly, you can plot for other victim rates. 04:13.730 --> 04:18.450 And here we will get all other segments which are running parallel to parallel data. 04:18.490 --> 04:18.930 Seven. 04:19.250 --> 04:22.810 So in case of XU2 where we have all the list of nets. 04:23.090 --> 04:25.370 And similarly for u seven. 04:26.530 --> 04:32.970 So by selecting each net you can verify how much crosstalk is there because of this aggression net. 04:34.210 --> 04:37.770 Another way to verify is just hover your cursor there. 04:38.210 --> 04:41.530 So it will give you all the information for each segment. 04:42.570 --> 04:48.130 So if you want me to create a detailed video of step by step process, how I have done this simulation 04:48.130 --> 04:51.290 for any layout, let me know in the comment section. 04:52.010 --> 04:55.890 Now we are good to go to discuss about near end and far end crosstalk. 05:00.530 --> 05:06.540 Next let's talk about near end and far end crosstalk between two parallel running active end coordinates. 05:07.020 --> 05:10.900 So let's assume a setup where we have active net and quiet net. 05:11.140 --> 05:15.980 And here both ends of the quiet net are terminated to avoid any reflection on the line. 05:16.180 --> 05:21.700 And similarly for active net, the other end of the signal is terminated to avoid reflection on that. 05:22.980 --> 05:29.500 So in this setup the measured noise voltage has a very different pattern for the noise in both ends 05:29.500 --> 05:30.780 of the victim nets. 05:31.540 --> 05:38.460 So to distinguish between these two ends, we label the end nearest to the source as near end of the 05:38.460 --> 05:44.340 victim net and the noise induced here we call it near end crosstalk. 05:45.140 --> 05:50.860 And similarly the end farthest from the source we call it far end of the victim net. 05:51.340 --> 05:55.260 And the noise induced there we call it far end crosstalk. 05:56.180 --> 06:03.110 Similarly we can define near end and far end in terms of direction of signal traveling Far end will 06:03.110 --> 06:08.990 be at the forward direction of signal propagation and near end as the backward direction of signal propagation. 06:14.350 --> 06:19.710 Let's talk about probing for near end crosstalk and far end crosstalk on victim nets. 06:19.950 --> 06:26.270 So when we connect oscilloscope on near end to observe the crosstalk we'll see a pattern like this. 06:26.950 --> 06:33.590 The near end noise rise up quickly and stays there for few nanoseconds and return back to zero. 06:34.630 --> 06:42.230 So on the active net, if we are sending 200 millivolts of signal at quiet net, the level is approx 06:42.230 --> 06:47.510 13 millivolts and which is about 6.5% of total signal level. 06:47.990 --> 06:52.470 On the other side at the far end will get really strange response. 06:52.590 --> 07:00.150 So if we connect crow and observe the noise for same active net parameters, the noise on quiet net 07:00.190 --> 07:06.360 will be in different shape and its amplitude will be approx 16 millivolts. 07:06.360 --> 07:12.560 So there will see a sharp dip which is 30% of the signal level on active net. 07:17.400 --> 07:22.440 Now I'm going to tell you some equations to estimate near end crosstalk and far end crosstalk. 07:23.000 --> 07:29.600 We can estimate the ratio of voltage on quiet line and voltage on signal line in terms of mutual capacitance 07:29.600 --> 07:30.760 and mutual inductance. 07:31.800 --> 07:37.800 So the equation will be voltage on quiet line divided by voltage of signal on active track is equal 07:37.800 --> 07:43.880 to one by four CML divided by CL plus LX divided by LX. 07:44.240 --> 07:52.040 So here CML and Llml are mutual capacitance and mutual inductance per unit length of coupled transmission 07:52.040 --> 07:52.440 line. 07:53.240 --> 07:59.200 And CL and LX are as you know, it's a capacitance and inductance per unit length of active line. 08:00.080 --> 08:06.380 So the ratio of voltage on quiet line divided by voltage on signal line of active net. 08:06.420 --> 08:14.220 We represent it with KB, which is known as backward coefficient or Near-end crosstalk coefficient. 08:14.820 --> 08:20.660 And similarly you can estimate the far end crosstalk coefficient, which is represented by KF, and 08:20.660 --> 08:27.900 it is the ratio of voltage at far end of quiet line and voltage at signal line of active net. 08:28.300 --> 08:30.860 It will be equal to one upon two v. 08:30.860 --> 08:39.860 So here v is the speed of signal on active net multiplied by CML divided by CL minus L divided by L. 08:40.820 --> 08:45.820 So these are mutual capacitance and mutual inductance and capacitance and inductance per unit length. 08:46.180 --> 08:48.740 And here V is the speed of signal on active net. 08:49.300 --> 08:53.620 So this is how you can estimate the far end crosstalk coefficient. 08:53.780 --> 08:56.500 And from that you can calculate the far end crosstalk. 09:01.620 --> 09:02.230 All right. 09:02.390 --> 09:07.990 So now I'm going to summarize this video by telling you couple of ways or methods we can use to reduce 09:07.990 --> 09:09.910 crosstalk between parallel running tracks. 09:10.150 --> 09:14.030 And the first one is you have to increase the pitch between signal tracks. 09:14.430 --> 09:22.150 Second is you have to use solid planes without any cutouts on the return path to avoid any ground bounce. 09:23.550 --> 09:26.150 Third one is keep coupled length short. 09:26.510 --> 09:29.230 So there will be very minimum crosstalk between those. 09:30.350 --> 09:34.270 Fourth is you have to route on strip line layers if you can. 09:34.670 --> 09:41.630 Fifth is do not share return pins in package or connectors because it will also lead to crosstalk. 09:42.350 --> 09:48.470 And last but not the least in case of high isolation required route tracks on different layers. 09:49.390 --> 09:50.550 So that's it for this video. 09:50.550 --> 09:56.030 In case you want me to create a video where I cover step by step process of simulations, how I have 09:56.030 --> 09:59.790 done the simulation on Aurora 17.4. 09:59.830 --> 10:01.350 Let me know in the comment section. 10:01.630 --> 10:02.830 See you in the next video.